When I develop a raw file to black and white, I usually use color calibration, but without the presets, placing it after the last colormodule in the pipeline. I take a two steps approach: the channel mix in grey-tones Tab gives me the basic look I want. In a second instance I then use the brightness-tab to get the contrast right.
My question: if the second instance of color-calibration uses the output of the first instance, why does it still handle rgb-channels, although there should not be any color information left? I could probably do both steps in one instance, but I like to keep my processing clean and simple.
Not the most significant question, but I don’t understand what happens.
Because an achromatic pixel is simply one where the channels are equal to each other (so R=G=B). A black pixel is thus [0, 0, 0] with some gray perhaps being [0.4, 0.4, 0.4] and display-referred white [1, 1, 1] (in the scene-referred part of the pixelpipe [1, 1, 1] has no particular meaning and is just another shade of gray).
Darktable doesn’t have a notion of “monochrome images”, only of images with three colour channels. “Monochrome” is just an interface notion, in a module that also handles changes which require three channels.
And, some may want to tint the monochrome image, or even use split toning. That requires three channels. So why make code more complex for no benefit?
EDIT: I should have use “gray-scale” instead of “monochrome” here: the gray tab of colour calibration creates a gray-scale image. If you then tone that in e.g. sepia, you don’t have a gray-scale image anymore, but you still have a monochrome image (in which you need the three colour channels for the sepia colour!)
@s7habo Has a couple of great videos on BW editing. They are worth watching if you have not seen them… and it might address some aspects of your question
Software that does black-n-white conversion can either 1) turn it into a single-channel image, or 2) retain the RGB channels, just make each pixel value R=G=B. indeed, the color information is gone either way, but the second way retains the three channels which can then be used to do things like introduce a color tone, like a brown for simulating sepia toning. if you do that, the image again becomes a “color” image, RGB not equal, this time with your sepia toning.
I can only speak with surety about my software, but I do this sort of toning by adding a control-point tone curve after the black/white conversion, selecting the blue channel, and pulling the bottom-left point to the right just a bit.
Thanks you for the advice. I have watched them a while ago and in one episode he explains, why he relocates the black and white conversion step in the pipeline after the other active Color modules, the reason being, that otherwise every color-module after color-calibration would only have a black and white output to work with. I will re-watch those videos, but I’m pretty sure, that’s what he said. Anyway, the multiple comments in my question are enough to answers my question and learn yet another detail about color management in dt, which is much appreciated.
You have misunderstood something then. The reason for putting the conversion instance of CC after color balance rgb is because the result of the b&w conversion is dependent on the existing colors. So before doing the conversion, he manipulates the colors so as to get the result he wants. Using color modules after the conversion will simply result in various color casts (which is something you want if doing sepia toning, for example).
Yes, I understand that now. However, I’m pretty confident, in the video he says otherwise. Thank you for the explanation. Now ist makes better sense to me, why relocating b/w conversion to the end of the color modules can still BE a reasonable Thing to do.